The Pharmacogenomics Journal (2011) 11, 337–347 & 2011 Macmillan Publishers Limited. All rights reserved 1470-269X/11 www.nature.com/tpj ORIGINAL ARTICLE

Polymorphic human H synthase-2 proteins and their interactions with substrates and inhibitors

W Liu1, EM Poole2,3,4, The cyclooxygenase (COX) activity of prostaglandin H synthase-2 (PGHS-2) 2,5 1 is implicated in colorectal cancer and is targeted by nonsteroidal anti- CM Ulrich and RJ Kulmacz inflammatory drugs (NSAIDs) and dietary nÀ3 fatty acids. We used purified, 1Department of Internal Medicine, University of recombinant proteins to evaluate the functional impacts of the R228H, Texas Health Science Center at Houston, E488G, V511A and G587R PGHS-2 polymorphisms on COX activity, fatty Houston, TX, USA; 2Public Health Sciences acid selectivity and NSAID actions. Compared to wild-type PGHS-2, COX Division, Fred Hutchinson Cancer Research activity with arachidonate was B20% lower in 488G and B20% higher in Center, Seattle, WA, USA; 3Department of Medicine, Brigham and Women’s Hospital, 511A. All variants showed time-dependent inhibition by the COX-2-specific Harvard Medical School, Boston, MA, USA; inhibitor (coxib) , but 488G and 511A had 30–60% higher 4Department of Epidemiology, Harvard School of residual COX activity; 511A also showed up to 70% higher residual activity 5 Public Health, Boston, MA, USA and Division of with other time-dependent inhibitors. In addition, 488G and 511A differed Preventive Oncology, German Cancer Research Center, Heidelberg, Germany significantly from wild type in Vmax values with the two fatty acids: 488G showed B20% less and 511A showed B20% more discrimination against Correspondence: eicosapentaenoic acid. The Vmax value for eicosapentaenoate was not Dr RJ Kulmacz, Department of Internal affected in 228H or 587R, nor were the Km values or the COX activation Medicine, University of Texas Health Science efficiency (with arachidonate) significantly altered in any variant. Thus, the Center at Houston, 6431 Fannin Street, Houston, TX 77030, USA. E488G and V511A PGHS-2 polymorphisms may predict who will most likely E-mail: [email protected] benefit from interventions with some NSAIDs or nÀ3 fatty acids. The Pharmacogenomics Journal (2011) 11, 337–347; doi:10.1038/tpj.2010.49; published online 15 June 2010

Keywords: prostaglandin H synthase-2 polymorphisms; cyclooxygenase; nonsteroidal anti- inflammatory drugs; ; eicosapentaenoic acid

Introduction

The cyclooxygenase (COX) activity of prostaglandin H synthase isoforms 1 and 2 (PGHS-1 and PGHS-2) catalyzes a key step in the biosynthesis of from nÀ6 and nÀ3 polyunsaturated fatty acids such as arachidonic acid (AA) and eicosapentaenoic acid (EPA).1 The prostaglandins, which include prosta- glandin E, and thromboxane, serve as potent signaling molecules in many pathophysiological processes.2 The two PGHS isoforms are functionally specialized, with PGHS-1 generally considered a constitutive, housekeeping , whereas PGHS-2 is strongly induced by cytokines and mitogens in a variety of cells and has been linked to the initiation and resolution phases of inflammation and to cell proliferation.3,4 PGHS-2 levels are elevated in many tumors and COX-2 activity has been implicated in the development of several types of cancer, including colorectal cancer.5,6 Received 31 October 2009; revised 16 March 2010; accepted 26 April 2010; published COX catalysis is strongly regulated by the availability of unesterified fatty acid online 15 June 2010 substrate, the structure of the substrate (for example, nÀ6 vs. nÀ3) and the level of cyclooxygenase-2 polymorphisms W Liu et al 338

of peroxide, which is required to generate the protein-based stocks of the fatty acids were dissolved in toluene, butylated radical at Tyr385 that begins the catalytic cycle.1 In hydroxytoluene (0.001 mol/mol) was added and the solu- addition, COX catalysis is the primary target of NSAIDs tions stored at À20 1C. Working suspensions (3–30 mM)in (nonsteroidal anti-inflammatory drugs); these agents 0.1 M Tris (pH 8.5) were prepared fresh each day and kept on include , , and the COX-2 ice. Tween 20 was from Anatrace (Maumee, OH, USA). inhibitors (coxibs), which are selective for inhibition of Coxibs were from Cayman Chemical (Ann Arbor, MI, USA) COX-2.3 NSAID use is linked to significantly decreased risk or Tocris Bioscience (Ellisville, MO, USA). Restriction of colorectal cancer in humans.5 and T4 DNA were purchased from New A limited number of nonsynonymous polymorphisms England Biolabs (Beverly, MA, USA). Oligonucleotides were have been identified in PTGS2, the that encodes from Integrated DNA Technologies (Coralville, IA, USA). human PGHS-2 (hPGHS-2); these polymorphisms include Reagents for DNA manipulation were from Promega R228H, E488G, V511A and G587R.7,8 The prevalence of (Madison, WI, USA). Plasmid transfer vector pAcSG2 and these polymorphisms is low, with a minor allele frequency BaculoGold linearized baculovirus DNA were from ranging from 0.8 to 5% in the populations studied, Pharmingen (San Diego, CA, USA). QuikChange single and suggesting that there is strong selective pressure on PTGS2. multisite-directed mutagenesis kits and Escherichia coli strain Epidemiological studies have examined the relation of the XL-10 were from Stratagene (La Jolla, CA, USA). Sf9 insect V511A polymorphism to the risks of colorectal adenoma or cells (from Spodoptera frugiperda), E. coli strain DH5a, Grace’s cancer,8–10 breast cancer11 and cardiovascular disease.12 supplemented medium and fetal bovine serum were from Because these studies were generally small (NB400–1000) Invitrogen (Carlsbad, CA, USA). Ni-NTA agarose was and the variant allele is rare (B4%), none of the studies purchased from Qiagen (Valencia, CA, USA). The integrity found a statistically significant association with risk. How- of each plasmid described below was verified by restriction ever, two of the colorectal cancer studies reported inverse enzyme digestion and by DNA sequencing at the Micro- associations with the A allele and some suggestion of an biology and Molecular Genetics Core Facility, UT Health NSAID interaction.8,10 Similarly, gene–NSAID interactions Science Center at Houston. All other reagents were obtained and pharmacogenetic relationships have been reported from Sigma (St Louis, MO, USA). in noncoding PTGS2 single-nucleotide polymorphisms (SNPs).13 as well as in other related to prostaglandin Construction of plasmid for recombinant, 6 Â His-tagged hPGHS-2 synthesis.14,15 The 6 Â His tag sequence was introduced after the signal There have been few published studies on the nonsynon- peptide cleavage site near the N terminus of the hPGHS-2 ymous polymorphisms listed above,7–11,16,17 and the V511A wild type (major allele at all targeted positions), as polymorphism is the only nonsynonymous PTGS2 variant previously reported.22 Nucleotides coding for the six that has been studied for functional effects so far. Fritsche histidine residues were inserted into the hPGHS-2/pAcSG2 et al.7 observed no major differences between wild-type vector using PCR with PfuUltra HF DNA polymerase hPGHS-2 and the V511A variant for of AA or (Stratagene) and the following 50-primers: sense, 50-pCAC , nor any differences between the alleles for CATCACCCTTGCTGTTCCCACCCATGTCAAAACC-30; anti- inhibition by indomethacin or . Lin et al.8 reported sense, 50-pATGGTGATGATTTGCTGTATGGCTGAGCGCCA that the major and minor alleles had similar V and K max m GGACC-30. values, but noted it was premature to conclude that the The PCR mixture was digested with DpnI; the resulting V511A variant had no functional effect without a more 7.4 kb product was separated by agarose gel electrophoresis thorough enzymological analysis. Several important aspects and ligated with T4 DNA ligase. The resulting plasmid was of COX-2 catalysis were not examined in the earlier studies. transformed into E. coli strain DH5a, and ampicillin- These aspects include: discrimination between nÀ6 and nÀ3 resistant colonies were analyzed by restriction enzyme substrates, which has considerable potential for diet-driven digest and DNA sequencing. modulation of signaling;18 peroxide-dependent feedback activation, which contributes to regulation of COX activity;19 and time-dependent COX inhibition effects, Construction of plasmid for hPGHS-2 with R228H, E488G, V511A which characterize the more potent pharmacological or G587R minor alleles agents.20 The present studies were undertaken to address Plasmids coding for hPGHS-2 with the desired point these important aspects of COX functionality in V511A and mutations were constructed with the QuikChange mutagen- three other variants of hPGHS-2 using the purified recombi- esis kit using the pAcSG2 plasmid containing the 6 Â His- nant proteins. tagged, wild-type hPGHS-2 cDNA as template. Primer pairs were (base changes underlined) as follows: R228H-f, 50-C TGGCTAGACAGCATAAACTGCGCCTTTTC-30; R228H-r, 50-G Materials and methods AAAAGGCGCAGTTTATGCTGTCTAGCCAG-30;E488G-f,50-C ATCGATGCTGTGGGGCTGTATCCTGCCC-30; E488G-r, 50-GG High-purity AA and EPA were obtained in sealed ampules GCAGGATACAGCCCCACAGCATCGATG-30; V511A-f, 50-GA from Nu Chek Prep (Elysian, MN, USA) and handled under AACCATGGTAGAAGCTGGAGCACCATTCTC-30; V511A-r, 50-G conditions found to minimize peroxide formation.21 Briefly, AGAATGGTGCTCCAGCTTCTACCATGGTTTC-30; G587R-f,

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 339

50-AGTTCTTCCCGCTCCCGACTAGATGATATC-30; G587R-r, min) by the reaction extent at complete self-inactivation 0 0 32 5 -GATATCATCTAGTCGGGAGCGGGAAGAACT-3 . (nmol O2).

Baculovirus generation and expression of recombinant proteins COX activation efficiency Generation, amplification and titer determination of Feedback activation by peroxide was assessed by measuring recombinant baculovirus containing the desired hPGHS-2 the COX activity of a fixed amount of enzyme (45–55 U) in cDNA, and expression of the recombinant proteins used the presence of varying amounts of GPx-1.23 Data for the 23 previously published techniques; procedures for expres- decline in the fraction of expressed COX activity with 24 sion of hPGHS-2 have been described. increasing peroxide scavenger were fitted to a straight line to obtain the slope; the slope obtained for hPGHS-2 wild type Purification of recombinant proteins in a given experiment was divided by the slope obtained for Purification of hPGHS-2 lacking a His tag has been hPGHS-2 with each minor allele to calculate the relative 24 described. The purification procedure for His-tagged activation efficiency for that variant. hPGHS-2 wild-type and mutant proteins was modified from a published protocol.25 The Sf9 cells expressing the recom- COX inhibition binant protein were suspended in 25 mM NaP ,20mM i COX inhibition kinetics were assessed at 30 1C with a imidazole, 1.0 mM phenol (pH 7.4) and sonicated; the ‘Standard’ electrode membrane in the standard reaction homogenate was centrifuged at 150 000 g at 4 1C for 1 h. mixture. Two experimental designs were used. The first The resulting pellet was resuspended with a Dounce focused on the forward reactions and involved timed homogenizer in 100 ml of 25 mM NaP , 100 mM NaCl, i preincubation of enzyme with inhibitor before the COX 20 mM imidazole, 0.1 mM phenol (pH 7.4), 1% v/v Tween reaction was started by addition of AA. In this design, the 20 and stirred for 1 h at 4 1C. Unextracted membrane residue enzyme (at B10À8 M) was preincubated with the desired was removed by centrifugation. The rest of the purification level of inhibitor in the reaction cuvette for various times was as described,25 except that buffers were supplemented before injection of 100 mM AA to determine the amount of with 0.1 mM phenol. Holoenzyme was reconstituted by enzyme that had not progressed to tight inhibitor–protein addition of .26 complexes. Inhibitors were dissolved in ethanol; the final ethanol concentration was below 1% and did not affect Protein characterization activity. AA was suspended at 30 mM in 0.1 M Tris-HCl (pH Total protein concentrations were determined by a modified 8.5). The second experimental design focused on the Lowry method.27 hPGHS-2 mutants were analyzed by kinetics of conversion of tight inhibitor–protein complexes polyacrylamide gel electrophoresis under denaturing condi- to looser complexes and dissociation of inhibitor. For this, tions,28 visualization with Coomassie staining of the gel and the enzyme (B10À5 M) was first incubated with a small densitometry using ImageJ29 to determine the fraction of excess of inhibitor (50 mM) to form an equilibrium mixture of total protein present as recombinant hPGHS-2. tight and looser inhibitor complexes. Aliquots of this mixture were then diluted 1500-fold into the reaction COX activity cuvette and re-equilibrated for various lengths of time Oxygen uptake was measured polarographically in standard before the amount of enzyme in looser inhibitor complexes assays at 30 1C in 100 mM KP (pH 7.2) with 1.0 mM phenol, i or without inhibitor bound was determined by addition of 1.0 mM heme and 0.025% Tween 20, using a ‘Standard’ 100 mM AA. electrode membrane (YSI, Yellow Springs, OH, USA).30 One unit of COX activity has an optimal velocity of 1 nmol O2 per min. The specific activity of each purified protein Peroxidase activity preparation was calculated by dividing the observed COX The stirred assay mixture contained 2.5 ml of 0.10 M Tris-HCl activity by the concentration of recombinant hPGHS-2 (pH 8.0) with 0.50 mM guaiacol, 1.0 mM heme and 0.40 mM HOOH at room temperature.33 Reaction was started by present. COX Km and Vmax values were determined by measuring the activity with 2–100 mM fatty acid at 25 1C with injection of enzyme and the initial velocity calculated from a YSI ‘High Sensitivity’ electrode membrane, in the presence the rate of increase in A436 due to oxidized guaiacol À1 À1 of 100 mM HOOH; these assay conditions ensure full (e436 ¼ 6.39 (mM oxidized guaiacol) cm ). One unit of activation and minimize dampening effects from the peroxidase (POX) activity has a velocity of 1 nmol guaiacol electrode membrane.31 The velocities were normalized to oxidized per min. those for the same batch of enzyme with 100 mM AA without added peroxide and fitted to the Michaelis–Menten equa- Statistical analysis tion using KaleidaGraph (Synergy Software, Reading, PA, The results of replicate experiments are presented as

USA) to estimate the Vmax and Km values. mean±s.d. of the mean. In most cases an unpaired, two- tailed t-test was used to assess statistical differences between COX self-inactivation individual variants and wild-type PGHS-2 (GraphPad COX self-inactivation rates for reactions in 100 mM AA were Software, La Jolla, CA, USA). A P-value below 0.05 was calculated by dividing the optimal velocity (nmol O2 per considered statistically significant.

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 340

Results measured at 25 1C with a ‘High-sensitivity’ oxygen electrode membrane to minimize dampening effects that can Expression and purification significantly attenuate velocity values. HOOH (100 mM) was Several expression runs, totaling 3–4 l of baculovirus- included in each reaction to ensure full COX activation, infected Sf9 cell culture, were pooled to prepare homogenates avoiding complications due to partial COX activation; this for purification of each recombinant protein. Wild-type is particularly important for reactions with EPA where hPGHS-2 and the four polymorphic proteins expressed well artifactually low COX velocities occur in the absence of in the baculovirus system, as assessed by the yield of COX exogenous peroxide.31 Batch-to-batch variations in expres- activity in the insect cell homogenate (Table 1). All of the sion and purification of recombinant PGHS-2 protein can proteins were readily solubilized by Tween 20 and purified lead to differences in COX-specific activity and confound to 75–90% electrophoretic homogeneity, with 38–64% comparison of COX activity in wild-type and mutant recovery of COX activity (Table 1). The COX-specific proteins. To control for such variations in specific activity, activities of the purified enzymes, calculated based on the we normalized the COX velocities for a particular protein amount of recombinant PGHS protein determined by assay reacted with a range of AA or EPA levels in the presence of of total protein and electrophoretic analysis, are shown in HOOH to the average velocity of the same protein on the Table 1. The specific activities of the 228H, 488G and 587R same day in replicate reactions with 100 mM AA without variants were 19–30% lower than the wild type, whereas the exogenous peroxide. This level of AA produces full COX value for the 511A variant was 55% higher than the wild activation without added peroxide. type. With the exception of the 511A variant, these specific Saturable responses of COX-2 reaction rate to AA and EPA activities fall in the range of 15–30 kU mgÀ1 observed for concentration were observed for wild-type PGHS-2 and each 31,34,35 purified PGHS-2 in our laboratory. To focus on of the variants, so values of Vmax and Km were estimated by possible effects of the polymorphic alleles that are specific fitting individual data sets to the Michaelis–Menten equa- for the COX activity, we normalized the COX activity for tion (examples are shown in Supplementary Figures S1–S5). each of the purified proteins to the corresponding POX The average Vmax values with AA for wild-type and activity determined on the same day (Table 1). The COX/ polymorphic hPGHS-2 proteins were all near unity (Figure POX ratio was 2.7 for wild-type PGHS-2, and the values for 1a), as expected because the rate with saturating AA was the 228H and 587R variants were not significantly different. used for normalization. The average Km value for wild-type In contrast, the COX/POX ratio for the 511A variant was hPGHS-2 was 2.1 mM AA (Figure 1b), quite comparable to the 20% higher, and the ratio for 488G B20% lower, than the value of 1.7±0.4 mM reported under the same reaction wild-type value; the difference was statistically significant conditions for hPGHS-2 without a His tag.31 None of the for both of these variants. Thus, the COX-specific activity mutations produced a significant difference in the average and the COX/POX ratio were both elevated for 511A and Km for AA or for EPA (Figure 1b). The average Vmax value for both decreased for 488G (Table 1). It is therefore likely that wild-type hPGHS-2 with EPA was 0.78, or almost 80% of the these two variant alleles have direct or indirect effects on Vmax with AA (Figure 1a). This shows that the fully activated COX catalytic efficiency. enzyme oxygenates EPA more slowly than AA. The average

Vmax values with EPA for the 228H and 587R variants were

COX Vmax and Km parameters indistinguishable from the wild-type value (Figure 1a). The COX kinetics of wild-type hPGHS-2 and the four However, the average Vmax value with EPA for 488G was variants were evaluated with prototypical substrates from 0.94, significantly higher than wild type and approaching the nÀ6 (AA) and nÀ3 (EPA) families, with precautions taken the Vmax value obtained with AA, and 511A had an average to ensure reliable COX activity measurements reflecting the Vmax value of 0.65 with EPA, significantly less than the wild- dependence on substrate concentration.31 COX activity was type protein (Figure 1a). Thus, the 488G variant led to less,

Table 1 Expression, recovery and enzymatic activities of recombinant hPGHS-2 proteins hPGHS-2 COX in homogenate Yield of COX after Electrophoretic COX-specific activityb COX/POXb constructa (kU/3 l culture) purification (%) homogeneity (%) (kU mgÀ1 PGHS-2) (U/U)

Wild typec 440 64 87 30±1 2.7±0.1 R228H 480 38 87 22±0 2.5±0.1 (P ¼ 0.07) E488G 300 46 75 21±1 2.1±0.1 (P ¼ 0.002) V511A 600 42 86 46±1 3.2±0.2 (P ¼ 0.018) G587R 550 52 90 24±2 2.5±0.1 (P ¼ 0.07) aVariant alleles are underlined. bMean±s.d. for triplicate determinations in standard assays. cMajor allele at all targeted residues.

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 341

Table 2 COX activation efficiencies of purified hPGHS-2 proteins with AA as substrate

hPGHS-2 constructa Activation efficiencyb (relative to wild type)

Wild typec 1.0 R228H 1.2±0.2 (n ¼ 3) P ¼ 0.31 E488G 1.2±0.2 (n ¼ 4) P ¼ 0.19 V511A 1.0±0.1 (n ¼ 4) P ¼ 0.65 G587R 1.0±0.1 (n ¼ 2) P ¼ 0.50

aVariant alleles are underlined. bValues represent the average±s.d. for the indicated number of determinations; P-values were determined by one-sample t-test. cMajor allele at all targeted residues.

Scheme 1 Mechanistic scheme for interaction of PGHS (E) with COX inhibitor (I).

the 587R variants was the same as the wild type. Conversely, both 228H and 488G had slightly higher activation efficiencies, though the differences were not statistically significant.

COX self-inactivation The rate constant for self-inactivation was 2.5 minÀ1 for wild-type hPGHS-2 and the values for the variants ranged from 2.5 to 2.8 minÀ1, indicating that none of the variants caused a major change in the COX self-inactivation process.

COX inhibition kinetics We analyzed the interactions of wild-type hPGHS-2 and the variant proteins with nimesulide, a COX-2-specific inhibitor that follows competitive, time-dependent kinetics and has proven useful as a probe of structure.36 The action of such inhibitors can be described by a two-step mechanism Figure 1 COX kinetic parameters for PGHS-2 wild type (WT) and such as that shown in Scheme 1. The initial binding of variants with AA (open circles) and EPA (filled circles). Values for Vmax relative to the rate with 100 mM AA and no HOOH (a) and Km (b) inhibitor (I) to the COX site of enzyme (E) to form a lower represent the average ±s.d. for triplicate determinations. Asterisks affinity complex (EI) is governed by a forward rate constant indicate values significantly different from wild type (Po0.05). (k1) and a reverse rate constant (kÀ1). This first interaction with inhibitor typically equilibrates very rapidly, and a B À5 typical Ki value (kÀ1/k1)is 10 M. In the case of time- dependent inhibitors, the EI complex undergoes a relatively slow transition to a higher affinity complex (EI0), governed and the 511A variant led to more, COX catalytic discrimina- 0 by rate constant k2. Unless a covalent bond is formed in EI , tion against EPA relative to AA. the complex can revert back to EI by the kÀ2 process. Two types of experimental approach were used to analyze COX activation efficiency the interactions of wild-type hPGHS-2 and the variant Titrations with a peroxide scavenger, glutathione perox- proteins with coxibs. The first involves preincubation of idase, were used to quantitatively compare the COX enzyme with inhibitor in the absence of substrate for various activation efficiencies of the four variant PGHS-2 proteins lengths of time before injection of saturating level of AA to to the wild-type enzyme. The results, shown in Table 2, quantitate the surviving COX activity. With shorter pre- indicate that the COX activation efficiency of the 511A and incubation periods, this approach focuses on the rate of

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 342

progression to the EI0 complex, which is only slowly reversible and thus does not contribute to COX catalysis when AA is added. When the preincubation is long enough, this approach provides an indication of the distribution of enzyme between E/EI and EI0 at equilibrium. It is important to note that the full inhibitory potency of time-dependent agents, such as nimesulide against COX-2, cannot be reliably assessed from conventional IC50 measurements because the IC50 value is determined both by the affinity of the rapid, reversible initial binding (E þ I3EI; Scheme 1) and by the rates of the slower EI ) EI0 and EI ( EI0 steps. On the other hand, the residual activity at long preincubation times is a useful measure of the maximal inhibition possible with a given agent, or conversely, what fraction of COX activity escapes inhibition even after equilibration with the agent (roughly mimicking the extended exposures in vivo). Thus, the ratio of the residual activities for variant and wild type gives a practical, quantitative index of any changes in COX-2 leakiness toward a given inhibitor. Nimesulide was chosen for the screening because it has Figure 2 Kinetics of COX inhibition in hPGHS-2 wild type (WT) and been established as a competitive, time-dependent inhibitor variants during preincubation with 2 mM nimesulide. Aliquots of each 36 that is selective for PGHS-2, and it is readily available. protein (10À8 M) were preincubated with (filled symbols) or without Range-finding assays with wild-type hPGHS-2 at a variety of (open symbols) inhibitor for the indicated lengths of time before the nimesulide concentrations indicated that equilibration was surviving COX activity (i.e., enzyme not in the EI0 form) was assayed by ± reached in a reasonable length of time with 2 mM inhibitor, injection of 100 mM AA. The values represent averages s.d. for triplicate so this level was used for the comparison with the variants. measurements. The results of this comparison are shown in Figure 2. The COX activity of each of the proteins was stable for at least 3 min of preincubation in control reactions without in- Table 3 Residual COX activities of nimesulide complexes of hibitor. With nimesulide present, wild-type hPGHS-2 and hPGHS-2 and the four polymorphic variants the four variants all progressively lost COX activity with a increasing preincubation time until the surviving activity hPGHS-2 construct Residual COX activity b plateaued as equilibrium was reached. The plateau fraction (fraction of control) of surviving COX activity in the wild type was 0.17, and c ± essentially the same values were found with the 228H and Wild type (n ¼ 9) 0.17 0.01 R228H(n ¼ 9) 0.18±0.00 587R variants (Figure 2 and Table 3). The surviving COX E488G(n ¼ 9) 0.22±0.01d activity plateaued at a markedly higher level (0.27) for the V511A(n ¼ 6) 0.27±0.01d 511A variant. This is some 60% higher than with wild-type G587R(n ¼ 9) 0.18±0.01 hPGHS-2 and indicates that the structural change at residue a 511 in the variant shifts the equilibrium significantly away Variant alleles are underlined. b ± from the tight EI0 complex toward the looser EI complex, After equilibration with 2 mM nimesulide, the average fraction ( s.d.) of COX activity surviving in the plateau region was calculated for each protein from the weakening the COX inhibition. The plateau level of COX data in Figure 3. activity in the E488G protein equilibrated with 2 mM cMajor allele at all targeted residues. nimesulide was 0.22, also significantly above the wild-type dSignificantly different from wild type (Po0.05). value (Figure 2 and Table 3). Thus, formation of the EI0 complex was decreased in the 488G variant, though to about half the degree seen in the 511A variant. In all similar to that in Figure 2, whereas COX-2 inhibition by proteins examined, the initial decay in COX activity FR122047 and SC-560 was not time dependent (data not followed first-order kinetics (as indicated by the early linear shown). The residual COX-2 activities after equilibration drops in the diagnostic semi-log plot in Figure 2), as is with the inhibitors are presented in Table 4. With FR122047 expected from the mechanism shown in Scheme 1, but the and SC-560 the 511A variant was inhibited the same or decline was too rapid at this level of inhibitor for a detailed slightly more than wild-type enzyme. Conversely, the comparison of k2 values. residual activity for the V511A variant was 10–70% greater The same approach was used to assess the surviving COX than wild type for all but one of the time-dependent agents activities of the wild-type and 511A hPGHS-2 proteins after tested. The exception to this pattern was , which equilibration with 2 mM levels of seven other coxibs. CAY- completely inhibited both enzymes. 10404, , NS-398, and diclofenac exhi- The second experimental approach to comparing interac- bited time-dependent inhibition of COX-2 in semi-log plots tions of the hPGHS-2 wild type and the 488G and 511A

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 343

Table 4 Residual COX activities after equilibration of hPGHS-2 and the V511A variant with selected COX inhibitors

Inhibitor Wild-typea residual COX activity V511A residual COX activity V511A/wild type (fraction of control)b (fraction of control)b

FR122047 0.96±0.03 (n ¼ 5) 0.97±0.04 (n ¼ 5) 1.0 CAY-10404c 0.78±0.03 (n ¼ 6) 0.88±0.01d (n ¼ 8) 1.1 Tenidapc 0.45±0.01 (n ¼ 5) 0.60±0.01d (n ¼ 5) 1.3 SC-560 0.40±0.01 (n ¼ 12) 0.37±0.01d (n ¼ 10) 0.9 Nimesulidec 0.17±0.01 (n ¼ 9) 0.27±0.01d (n ¼ 6) 1.6 NS-398c 0.035±0.003 (n ¼ 6) 0.058±0.002d (n ¼ 6) 1.7 Valdecoxibc 0.035±0.003 (n ¼ 5) 0.042±0.001e (n ¼ 5) 1.2 Diclofenacc 0(n ¼ 2) 0 (n ¼ 2) — aMajor allele at all targeted residues. bThe values shown represent the averages from reactions with 2 mM inhibitor. cTime-dependent COX inhibition with PGHS-2. dPo0.0001. ePo0.002.

variants with inhibitor involved pre-equilibration of a concentrated solution of enzyme with a small excess of nimesulide to convert the enzyme to a mixture of EI and EI0 forms. This mixture was then diluted 1500-fold into buffer without substrate and preincubated for varying lengths of time to allow redistribution of the EI0 complex into EI and E forms. Finally, a saturating level of AA was added to assay the increase in EI and E. The results are presented in Figure 3, normalized to the corresponding reference enzyme/inhibitor mixture that was assayed immediately on dilution. After a lag of B25 s, the wild-type COX activity increased in a first-order manner, with a rate constant of 4.6 Â 10À3 sÀ1 and a maximal increase of 4.0-fold. The COX activity of the 488G variant increased at 3.5 Â 10À3 sÀ1, B25% slower than wild type, to eventually reach a 3.5-fold increase in activity. The increase in COX activity on dilution was B40% faster in the V511A variant (6.4 Â 10À3 sÀ1) than in wild type, reaching a 3.1-fold increase. The clear differences in the rate of recovery of COX activity on dilution of the enzyme–inhibitor mixtures indicate that redistribution from the EI0 complex is faster in the 511A variant, and slower in the 488G variant, compared to wild- Figure 3 Kinetics of re-equilibration of tightly bound nimesulide type PGHS-2. It is worth noting that the magnitude of the complexes of hPGHS-2 wild type (WT) and the E488G and V511A COX increase observed on dilution of the enzyme–nimesu- variants upon 1500-fold dilution. The diluted samples were subse- quently incubated for the indicated lengths of time before the COX lide mixtures decreased in the order, wild type 4488G activity was assayed by injection of 100 mM AA. Activities were 4511A, as might be expected from the opposite rank order normalized to control values obtained by dilution into buffer already of residual activity observed in the experiments in Figure 2. containing 100 mM AA. Averages±s.d. for two separate experiments are Overall, the results of the second approach (Figure 3) presented. Each set of data points was fitted to a single exponential corroborate the results from the first approach (Figure 2 equation with a 25 s lag. and Tables 3 and 4), and establish that the 488G and 511A variants alter the interaction of hPGHS-2 with one or more COX-2 selective inhibitors. effects on the COX activity with arachidonate, on the enzyme’s substrate selectivity and on the interactions with Discussion COX-2 selective inhibitors. These functional effects are in the 20–70% range, consistent with the selective pressures expected The results of this study show that two (488G and 511A) of for an enzyme that figures in several major pathophysiological the four hPGHS-2 variants tested have significant functional processes. The in vivo impacts of polymorphisms are more

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 344

readily evaluated when they result in large changes in enzyme FR122047, which showed only simple reversible inhibition kinetics. Nevertheless, there are a number of cases where of COX-2 (Table 4). physiological or clinical effects are linked to variants whose Crystallographic data for PGHS-247 show that Glu488 kinetic parameters after purification are at least 30% of the (Glu502 in ovine PGHS-1 numbering) is positioned at some wild-type Vmax and less than three times the wild-type Km. distance from the COX active site (Figure 4b). Thus, the Examples include glucose-6-phosphate dehydrogenase L382P perturbation of substrate selectivity and inhibitor inter- in Drosophila melanogaster,37,38 chicken amylase S,39,40 human actions observed in the 488G variant is not readily ubiquitin C-terminal L1 I93M,41 several variants of explained by local structural changes. However, the side human propionyl-CoA carboxylase,42 human serine hydro- chain of Glu488 appears to form a salt bridge with Arg442 xymethyl S394N,43 human P450 and to be hydrogen bonded to the side chain of Tyr481. A284P and V605F,44 and human coproporphyrinogen oxidase These interactions would be lost in the 488G variant, A814C.45 Given these precedents and the importance of COX- with the potential for long-range structural changes affect- 2 to prostanoid signaling pathways, it would seem worthwhile ing active site events, such as the EI3EI0 transition. In to evaluate more closely the impact of the V511A and E488G addition, Glu488 is just two residues away from Tyr490 variants on the effectiveness of therapeutic intervention with (Tyr504 in ovine PGHS-1 numbering; Figure 4b), the site of coxibs or elevated nÀ3 fatty acid intake. the other tyrosyl radical in PGHS-2.34 It will thus be of The V511A polymorphism has been previously examined interest to examine tyrosyl radical dynamics in the 488G for functional changes and no major differences with variant. wild type in COX kinetic parameters, product profile or Aspirin and coxibs are potent chemopreventive agents for inhibition by NSAID were reported.7,8 Despite this outcome, colorectal adenoma recurrence;51–55 aspirin has also been it was recognized that more enzymological analysis was shown to prevent colorectal cancer with about a 10-year needed to fully characterize possible functional impacts of delay.56 As PGHS-2 is a major target of NSAIDs and coxibs as the 511A allele.8 The present studies incorporate several well as an important mediator of inflammation, polymorph- significant methodological improvements for examination isms in PTGS2 have been extensively examined for their of V511A and three other PGHS-2 polymorphisms. These association with colorectal adenoma or cancer risk.8,9,13,57–71 include the use of purified recombinant proteins to avoid Although several studies have shown associations with host cell lipids that might modify the interactions with noncoding PTGS2 polymorphisms, few studies have evalu- hydrophobic substrates and inhibitors, quantitation of the ated potential SNP–NSAIDs interactions.57,59,60,63,67,70,71 The COX/POX ratio as a sensitive index of COX-selective effects, V511A polymorphism of PTGS2 has significant prevalence examination of substrate discrimination between the two only in the African-American population (thus far), and the major prostanoid precursors (AA and EPA) and use of E488G polymorphism appears to be very rare.7,8 The inhibitor assays focused on the time-dependent inhibition polymorphisms studied here exist with allele frequencies that is the major mode of action of selective coxibs.46 These between 0.8 and 5% in the African-American and Asian new approaches revealed some functional effects of the populations. In prior studies, the populations carrying a 511A variant that could not be detected in the earlier variant allele were up to 5%, suggesting that a significant studies. proportion of African Americans can experience different As depicted in Figure 4a, the V511A polymorphism is in a NSAID pharmacokinetics. Differing effects among men and segment that forms part of the COX active site.47 Although women were not observed in one study,61 but should be the side chain of residue 511 is not positioned to interact evaluated further. Nevertheless, these polymorphisms could with bound fatty acid or inhibitor in crystallographic have important pharmacogenetic implications, given the structures of PGHS-2,47 aVal) Ala substitution at this role of NSAIDs in colorectal adenoma prevention. Coxibs position could plausibly change the protein packing and (PGHS-2-specific inhibitors) are potent chemopreventive the position of bound substrate with respect to the catalytic agents for colorectal adenoma recurrence, but their use is radical at Tyr371 (Tyr385 in conventional ovine PGHS-1 associated with increased risk of and numbering), indirectly affecting the COX catalytic rate. AA other cardiovascular events.72 Evaluations of these poly- and EPA have differences in binding interactions in PGHS- morphisms in relation to coxibs may provide information 148 and the situation is likely to be similar in PGHS-2; such on who is most likely to benefit from coxib use and who differences might well underlie the differential impact of may be at risk of adverse events.6 Given that the incidence

V511A on the COX Vmax values with EPA and AA as of colorectal cancer is elevated in African Americans substrate (Figure 1a). Most coxibs compete with fatty acid compared to Caucasians73 and that Japan has one of the binding at the COX site,49 so the structural changes just highest colorectal cancer incidences in the world,74 further described can be invoked to explain the observed changes in study is required to determine whether nonsynonymous inhibitor interactions in the 511A variant. An important polymorphisms in PTGS2 can influence the chemo- structural determinant for COX-2 time-dependent inhibi- preventive efficacy of NSAIDs in populations that are likely tion has been identified just two residues away at Val509 to carry these polymorphisms. This information may be very (Figure 4a).36,50 This may have some connection to the useful for the tailoring of therapy in these populations so as observation that the V511A/wild-type differences tended to to minimize the serious adverse events associated with be larger with time-dependent coxibs than with SC-560 and NSAID use.6

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 345

Figure 4 Crystallographic structure of murine PGHS-2 (6cox.pdb), with views focused on Val511 (a) and Glu488 (b). Residue numbers are for human PGHS-2, with ovine PGHS-1 equivalents in parentheses. Other highlights include heme, bound inhibitor (SC-558), the tyrosyl radical sites at Tyr371(385) and Tyr490(504), the coxib specificity determinant at Val509(523), and the Arg442(456) and Tyr481(495) residues that interact with Glu488.

The 511A variant has a lower relative Vmax for EPA than activity for EPA compared to AA seen with PGHS-1 and to does the wild type (Figure 1a), indicating that this variant a lesser extent with PGHS-2.1 Accordingly, individuals discriminates more strongly against the nÀ3 substrate carrying the 511A variant of PTGS2 might be expected to relative to the nÀ6 substrate. Increased dietary intake of benefit more from dietary intervention designed to increase nÀ3 fatty acids such as EPA has been found to bring health the nÀ3/nÀ6 ratio in tissue lipids than would individuals benefits.75 One proposed mechanism is the lower COX with the wild-type allele. A recent study12 reported no

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 346

association between the V511A polymorphism and risk of 15 Poole EM, Bigler J, Whitton J, Sibert JG, Potter JD, Ulrich CM. chronic heart disease or ischemic , but dietary fatty and arachidonate 5-lipoxygenase polymorphisms acid intake was not examined as an independent variable. and risk of colorectal polyps. Cancer Epidemiol Biomarkers Prev 2006; 15: 502–508. In summary, our biochemical studies show that two 16 Choi JH, Park HS, Oh HB, Lee JH, Suh YJ, Park CS et al. - genetic variants in PGHS-2 can affect enzyme activity, related gene polymorphisms in ASA-intolerant asthma: an association substrate selectivity and inhibition with coxibs. Further with a haplotype of 5-lipoxygenase. Hum Genet 2004; 114: 337–344. studies are needed to evaluate whether these functional 17 Hamajima N, Takezaki T, Matsuo K, Saito T, Inoue M, Hirai T et al. Genotype frequencies of cyclooxygenease 2 (COX2) rare polymorph- differences affect the efficacy of dietary and pharmaco- isms for Japanese with and without colorectal cancer. Asian Pac J Cancer logical interventions, particularly in African Americans, the Prev 2001; 2: 57–62. population in which these two variants have been observed. 18 Wada M, DeLong CJ, Hong YH, Rieke CJ, Song I, Sidhu RS et al. Enzymes and receptors of prostaglandin pathways with arachidonic acid-derived versus eicosapentaenoic acid-derived substrates and products. J Biol Conflict of interest Chem 2007; 282: 22254–22266. 19 Kulmacz RJ. Regulation of cyclooxygenase catalysis by hydroperoxides. The authors declare no conflict of interest. Biochem Biophys Res Commun 2005; 338: 25–33. 20 Blobaum AL, Marnett LJ. Structural and functional basis of cyclooxy- genase inhibition. J Med Chem 2007; 50: 1425–1441. Acknowledgments 21 Chen W, Pawelek TR, Kulmacz RJ. Hydroperoxide dependence and cooperative cyclooxygenase kinetics in prostaglandin H synthase-1 and We thank Ms Liren Xiao for assistance with the statistical analysis. -2. J Biol Chem 1999; 274: 20301–20306. This work was financially supported by National Institutes of 22 Smith T, Leipprandt J, DeWitt D. Purification and characterization of the human recombinant histidine-tagged prostaglandin endoper- Health, USA (Grant CA 114467). oxide H synthases-1 and -2. Arch Biochem Biophys 2000; 375: 195–200. 23 Bambai B, Rogge CE, Stec B, Kulmacz RJ. Role of Asn-382 and Thr-383 References in activation and inactivation of human prostaglandin H synthase cyclooxygenase catalysis. J Biol Chem 2004; 279: 4084–4092. 1 Smith WL. Nutritionally essential fatty acids and biologically indis- 24 Kulmacz RJ, Wang LH. Comparison of hydroperoxide initiator require- pensable . Trends Biochem Sci 2008; 33: 27–37. ments for the cyclooxygenase activities of prostaglandin H synthase-1 2 Funk CD. Prostaglandins and : advances in and -2. J Biol Chem 1995; 270: 24019–24023. biology. Science 2001; 294: 1871–1875. 25 MirAfzali Z, Leipprandt JR, McCracken JL, DeWitt DL. Fast, efficient 3 Simmons DL, Botting RM, Hla T. Cyclooxygenase isozymes: the reconstitution of the cyclooxygenases into proteoliposomes. Arch biology of prostaglandin synthesis and inhibition. Pharmacol Rev Biochem Biophys 2005; 443: 60–65. 2004; 56: 387–437. 26 Tsai A-L, Wu G, Palmer G, Bambai B, Koehn JA, Marshall PJ et al. 4 Serhan CN, Yacoubian S, Yang R. Anti-inflammatory and proresolving Rapid kinetics of tyrosyl radical formation and heme redox state lipid mediators. Annu Rev Pathol 2008; 3: 279–312. changes in prostaglandin H synthase-1 and -2. J Biol Chem 1999; 274: 5 Mann JR, DuBois RN. Cyclooxygenase-2 and gastrointestinal cancer. 21695–21700. Cancer J 2004; 10: 145–152. 27 Peterson GL. Determination of total protein. Methods Enzymol 1983; 91: 6 Ulrich CM, Bigler J, Potter JD. Non-steroidal anti-inflammatory drugs for 95–119. cancer prevention: promise, perils and pharmacogenetics. Nat Rev 28 Laemmli UK. Cleavage of structural proteins during the assembly of the Cancer 2006; 6: 130–140. head of bacteriophage T4. Nature 1970; 227: 680–685. 7 Fritsche E, Baek SJ, King LM, Zeldin DC, Eling TE, Bell DA. Functional 29 Abramoff MD, Magelhaes PJ, Ram SJ. Image processing with ImageJ. characterization of cyclooxygenase-2 polymorphisms. J Pharmacol Exp Biophotonics Int 2004; 11: 36–42. Ther 2001; 299: 468–476. 30 Kulmacz RJ, Lands W. Cyclo-: measurement, purification and 8 Lin HJ, Lakkides KM, Keku TO, Reddy ST, Louie AD, Kau IH et al. properties. In: Benedetto C, McDonald-Gibson RG, Nigam S, Slater TF Prostaglandin H synthase 2 variant (Val511Ala) in African Americans (eds). Prostaglandins and Related Substances: A Practical Approach.IRL may reduce the risk for colorectal neoplasia. Cancer Epidemiol Biomarkers Press: Washington, DC, 1987, pp 209–227. Prev 2002; 11: 1305–1315. 31 Liu W, Cao D, Oh SF, Serhan CN, Kulmacz RJ. Divergent cyclooxy- 9 Goodman JE, Bowman ED, Chanock SJ, Alberg AJ, Harris CC. genase responses to fatty acid structure and peroxide level in fish and Arachidonate lipoxygenase (ALOX) and cyclooxygenase (COX) poly- mammalian prostaglandin H synthases. FASEB J 2006; 20: 1097–1108. morphisms and colon cancer risk. Carcinogenesis 2004; 25: 2467–2472. 32 Smith WL, Lands WE. Oxygenation of unsaturated fatty acids by 10 Sansbury LB, Bergen AW, Wanke KL, Yu B, Caporaso NE, Chatterjee N soybean lipoxygenase. J Biol Chem 1972; 247: 1038–1047. et al. Inflammatory cytokine gene polymorphisms, nonsteroidal anti- 33 Kulmacz RJ. Concerted loss of cyclooxygenase and peroxidase activities inflammatory drug use, and risk of adenoma polyp recurrence in from prostaglandin H synthase upon proteolytic attack. Prostaglandins the polyp prevention trial. Cancer Epidemiol Biomarkers Prev 2006; 15: 1989; 38: 277–288. 494–501. 34 Rogge CE, Liu W, Wu G, Wang LH, Kulmacz RJ, Tsai AL. Identification of 11 Moorman PG, Sesay J, Nwosu V, Kane JG, de Cotret AR, Worley K et al. Tyr504 as an alternative tyrosyl radical site in human prostaglandin H Cyclooxygenase 2 polymorphism (Val511Ala), nonsteroidal anti-inflam- synthase-2. Biochemistry 2004; 43: 1560–1568. matory drug use and breast cancer in African American women. Cancer 35 Rogge CE, Ho B, Liu W, Kulmacz RJ, Tsai AL. Role of Tyr348 in Tyr385 Epidemiol Biomarkers Prev 2005; 14: 3013–3014. radical dynamics and cyclooxygenase inhibitor interactions in prosta- 12 Lee CR, North KE, Bray MS, Couper DJ, Heiss G, Zeldin DC. glandin H synthase-2. Biochemistry 2006; 45: 523–532. Cyclooxygenase polymorphisms and risk of cardiovascular events: the 36 Guo Q, Wang LH, Ruan KH, Kulmacz RJ. Role of Val509 in time- Risk in Communities (ARIC) study. Clin Pharmacol Ther dependent inhibition of human prostaglandin H synthase-2 cycloox- 2008; 83: 52–60. ygenase activity by isoform-selective agents. J Biol Chem 1996; 271: 13 Ulrich CM, Whitton J, Yu JH, Sibert J, Sparks R, Potter JD et al. PTGS2 19134–19139. (COX-2) -765G 4 C promoter variant reduces risk of colorectal 37 Eanes WF, Katona L, Longtine M. Comparison of in vitro and in vivo adenoma among nonusers of nonsteroidal anti-inflammatory drugs. activities associated with the G6PD allozyme polymorphism in Cancer Epidemiol Biomarkers Prev 2005; 14: 616–619. Drosophila melanogaster. Genetics 1990; 125: 845–853. 14 Ulrich CM, Bigler J, Sparks R, Whitton J, Sibert JG, Goode EL et al. 38 Labate J, Eanes WF. Direct measurement of in vivo flux differences Polymorphisms in PTGS1 ( ¼ COX-1) and risk of colorectal polyps. between electrophoretic variants of G6PD from Drosophila melanoga- Cancer Epidemiol Biomarkers Prev 2004; 13: 889–893. ster. Genetics 1992; 132: 783–787.

The Pharmacogenomics Journal Enzyme kinetics of cyclooxygenase-2 polymorphisms W Liu et al 347

39 Gapusan RA, Yardley DG, Hughes BL. The amylase gene-enzyme system 58 Ali IU, Luke BT, Dean M, Greenwald P. Allellic variants in regulatory of chickens. II. Biochemical characterization of allozymes. Biochem Genet regions of cyclooxygenase-2: association with advanced colorectal 1990; 28: 553–560. adenoma. Br J Cancer 2005; 93: 953–959. 40 Hughes BL, Suniga RG, Yardley DG. Influence of amylase genotypes on 59 Cox DG, Pontes C, Guino E, Navarro M, Osorio A, Canzian F et al. growthrateandfeedconversionofchickens.Poult Sci 1994; 73: 953–957. Polymorphisms in prostaglandin synthase 2/cyclooxygenase 2 41 Nishikawa K, Li H, Kawamura R, Osaka H, Wang YL, Hara Y et al. (PTGS2/COX2) and risk of colorectal cancer. Br J Cancer 2004; 91: Alterations of structure and hydrolase activity of parkinsonism-asso- 339–343. ciated human ubiquitin carboxyl-terminal hydrolase L1 variants. 60 Pereira C, Pimentel-Nunes P, Brandao C, Moreira-Dias L, Medeiros R, Biochem Biophys Res Commun 2003; 304: 176–183. Dinis-Ribeiro M. COX-2 polymorphisms and colorectal cancer risk: a 42 Jiang H, Rao KS, Yee VC, Kraus JP. Characterization of four variant forms strategy for chemoprevention. Eur J Gastroenterol Hepatol 2010; 22: of human propionyl-CoA carboxylase expressed in Escherichia coli. J Biol 607–613. Chem 2005; 280: 27719–27727. 61 Sansbury LB, Millikan RC, Schroeder JC, North KE, Moorman PG, Keku 43 Fu TF, Hunt S, Schirch V, Safo MK, Chen BH. Properties of human and TO et al. COX-2 polymorphism, use of nonsteroidal anti-inflammatory rabbit cytosolic serine hydroxymethyltransferase are changed by single drugs, and risk of colon cancer in African Americans (United States). nucleotide polymorphic mutations. Arch Biochem Biophys 2005; 442: Cancer Causes Control 2006; 17: 257–266. 92–101. 62 Gunter MJ, Canzian F, Landi S, Chanock SJ, Sinha R, Rothman N. 44 Pandey AV. Biochemical analysis of mutations in P450 oxidoreductase. Inflammation-related gene polymorphisms and colorectal adenoma. Biochem Soc Trans 2006; 34(Part 6): 1186–1191. Cancer Epidemiol Biomarkers Prev 2006; 15: 1126–1131. 45 Li T, Woods JS. Cloning, expression, and biochemical properties of 63 Tan W, Wu J, Zhang X, Guo Y, Liu J, Sun T et al. Associations of CPOX4, a genetic variant of coproporphyrinogen oxidase that functional polymorphisms in cyclooxygenase-2 and 12-lipoxy- affects susceptibility to mercury toxicity in humans. Toxicol Sci 2009; genase with risk of occurrence and advanced disease status of colorectal 109: 228–236. cancer. Carcinogenesis 2007; 28: 1197–1201. 46 Copeland RA, Williams JM, Giannaras J, Nurnberg S, Covington M, 64 Hubner RA, Muir KR, Liu JF, Logan RF, Grainge MJ, Houlston RS. Pinto D et al. Mechanism of selective inhibition of the inducible isoform Polymorphisms in PTGS1, PTGS2 and IL-10 do not influence colorectal of prostaglandin G/H synthase. Proc Natl Acad Sci USA 1994; 91: adenoma recurrence in the context of a randomized aspirin interven- 11202–11206. tion trial. Int J Cancer 2007; 121: 2001–2004. 47 Kurumbail RG, Stevens AM, Gierse JK, McDonald JJ, Stegeman RA, Pak 65 Ueda N, Maehara Y, Tajima O, Tabata S, Wakabayashi K, Kono S. Genetic JY et al. Structural basis for selective inhibition of cyclooxygenase-2 by polymorphisms of cyclooxygenase-2 and colorectal adenoma risk: the Self anti-inflammatory agents. Nature 1996; 384: 644–648. Defense Forces Health Study. Cancer Sci 2008; 99: 576–581. 48 Malkowski MG, Thuresson ED, Lakkides KM, Rieke CJ, Micielli R, Smith 66 Xing LL, Wang ZN, Jiang L, Zhang Y, Xu YY, Li J et al. Cyclooxygenase 2 WL et al. Structure of eicosapentaenoic and linoleic acids in the polymorphism and colorectal cancer: -765G4C variant modifies risk cyclooxygenase site of prostaglandin endoperoxide H synthase-1. J Biol associated with smoking and body mass index. World J Gastroenterol Chem 2001; 276: 37547–37555. 2008; 14: 1785–1789. 49 Marnett LJ, Rowlinson SW, Goodwin DC, Kalgutkar AS, 67 Ashktorab H, Tsang S, Luke B, Sun Z, Adam-Campbell L, Kwagyan J Lanzo CA. Arachidonic acid oxygenation by COX-1 and COX-2. et al. Protective effect of Cox-2 allelic variants on risk of colorectal Mechanisms of catalysis and inhibition. J Biol Chem 1999; 274: adenoma development in African Americans. Anticancer Res 2008; 28: 22903–22906. 3119–3123. 50 Gierse JK, McDonald JJ, Hauser SD, Rangwala SH, Koboldt CM, Seibert 68 Gong Z, Bostick RM, Xie D, Hurley TG, Deng Z, Dixon DA et al. K. A single amino acid difference between cyclooxygenase-1 (COX-1) Genetic polymorphisms in the cyclooxygenase-1 and cyclooxygenase-2 and -2 (COX-2) reverses the selectivity of COX-2 specific inhibitors. J Biol genes and risk of colorectal adenoma. Int J Colorectal Dis 2009; 24: Chem 1996; 271: 15810–15814. 647–654. 51 Baron JA, Cole BF, Sandler RS, Haile RW, Ahnen D, Bresalier R et al. A 69 Thompson CL, Plummer SJ, Merkulova A, Cheng I, Tucker TC, Casey G randomized trial of aspirin to prevent colorectal adenomas. N Engl J Med et al. No association between cyclooxygenase-2 and uridine diphos- 2003; 348: 891–899. phate glucuronosyltransferase 1A6 genetic polymorphisms and colon 52 Sandler RS, Halabi S, Baron JA, Budinger S, Paskett E, Keresztes R et al. A cancer risk. World J Gastroenterol 2009; 15: 2240–2244. randomized trial of aspirin to prevent colorectal adenomas in 70 Barry EL, Sansbury LB, Grau MV, Ali IU, Tsang S, Munroe DJ et al. patients with previous colorectal cancer. N Engl J Med 2003; 348: Cyclooxygenase-2 polymorphisms, aspirin treatment, and risk for 883–890. colorectal adenoma recurrence—data from a randomized clinical trial. 53 Arber N, Eagle CJ, Spicak J, Racz I, Dite P, Hajer J et al. Celecoxib for the Cancer Epidemiol Biomarkers Prev 2009; 18: 2726–2733. prevention of colorectal adenomatous polyps. N Engl J Med 2006; 355: 71 Hoff JH, te Morsche RH, Roelofs HM, van der Logt EM, Nagengast FM, 885–895. Peters WH. COX-2 polymorphisms -765G–4C and -1195A–4G and 54 Baron JA, Sandler RS, Bresalier RS, Quan H, Riddell R, Lanas A et al. A colorectal cancer risk. World J Gastroenterol 2009; 15: 4561–4565. randomized trial of for the chemoprevention of colorectal 72 Bertagnolli MM. Chemoprevention of colorectal cancer with cyclooxy- adenomas. Gastroenterology 2006; 131: 1674–1682. genase-2 inhibitors: two steps forward, one step back. Lancet Oncol 55 Bertagnolli MM, Eagle CJ, Zauber AG, Redston M, Solomon SD, Kim K 2007; 8: 439–443. et al. Celecoxib for the prevention of sporadic colorectal adenomas. N 73 Horner MJ, Ries LAG, Krapcho M, Neyman N, Aminou R, Howlader N Engl J Med 2006; 355: 873–884. et al. (eds). SEER Cancer Statistics Review, 1975–2006.NationalCancer 56 Flossmann E, Rothwell PM. Effect of aspirin on long-term risk of Institute: Bethesda, MD. http://seer.cancer.gov/csr/1975_2006/, based on colorectal cancer: consistent evidence from randomised and observa- November 2008 SEER data submission, posted to the SEER web site, 2009. tional studies. Lancet 2007; 369: 1603–1613. 74 Center MM, Jemal A, Smith RA, Ward E. Worldwide variations in 57 Siezen CL, Bueno-de-Mesquita HB, Peeters PH, Kram NR, van Doeselaar colorectal cancer. CA Cancer J Clin 2009; 59: 366–378. M, van Kranen HJ. Polymorphisms in the genes involved in the 75 Ruxton CH, Reed SC, Simpson MJ, Millington KJ. The health benefits of arachidonic acid-pathway, fish consumption and the risk of colorectal omega-3 polyunsaturated fatty acids: a review of the evidence. J Hum cancer. Int J Cancer 2006; 119: 297–303. Nutr Diet 2004; 17: 449–459.

Supplementary Information accompanies the paper on The Pharmacogenomics Journal website (http://www.nature.com/tpj)

The Pharmacogenomics Journal